Evidence generation plan for artificial intelligence (AI) technologies for assessing and triaging skin lesions referred to the urgent suspected skin cancer pathway
6 Implementation considerations
When the technology is implemented, steps should be taken to mitigate the potential risk of missed or delayed cancer diagnoses when using DERM during the evidence generation period by:
-
doing a healthcare professional review for people with black or brown skin
-
regular monitoring of DERM's performance to maintain accuracy.
-
using additional protocols when necessary, such as:
-
a national governance framework to ensure local oversight of use of DERM
-
a healthcare professional review.
-
The company should work with NHS commissioners and providers to implement the technology in real-world settings. Planning for a prespecified period for the set-up of the technology is advised. During this period, training and implementation should be done before data collection is started, to account for learning effects. NHS England has set up an independent data oversight group that will support evidence generation.
An inclusive user-led design approach should be followed, with individuals across key stakeholder groups at implementation sites brought together at the start to map the pathway and plan for implementation.
Care should be taken to ensure that:
-
lesions in people with black or brown skin have a second read by a healthcare professional throughout the evidence generation period
-
people with disabilities are supported to access the teledermatology service
-
information is provided in:
-
clear, plain language to enable informed consent to be given
-
alternative languages, or local translation services are used as needed.
-
For the implementation of the real-world study, efforts should be made to select sites with a range of different characteristics to ensure generalisability of the results. Site characteristics could include:
-
type of centre, for example, community diagnostic centres versus hospital sites
-
racial or ethnic diversity of the population
-
image takers, for example, medical photographers versus healthcare assistants
-
local IT systems
-
local dermatologist workforce (for example, vacancy rate and seniority mix).
For safe implementation, when DERM is first introduced at a site, specialist dermatologists may wish to use the tool with a second read for an agreed number of interactions. This approach can be used to understand whether the technology can be deployed safely and what the influence on decision making would likely have been (for example, onward referrals). It may also collect some relevant data items (for example, test failure rate or number of indeterminate findings).
For people with black or brown skin (Fitzpatrick skin types 5 and 6), a second read by a healthcare professional should take place.
The company should provide training for staff in using the technology as well as ongoing support. To assess the potential for automation bias after deployment, the company may want to track the rate of diagnostic disagreement over time.
Potential barriers to implementation include:
-
hesitation of clinical teams to adopt new technology
-
the availability of NHS funding to cover the costs of implementing the technology in clinical practice
-
burden on clinical staff, including the need to have training ahead of implementation, data collection and follow-up
-
differences in dermatology pathways across the NHS
-
differences in skill and experience among staff when using the AI technology.
This page was last updated: